About the Workchart

What it's for

A senior leader asking "where will AI break my organization, and where will it create value?" needs a map. Today they have two tools — neither sufficient:

• The org chart describes reporting hierarchy. It tells you nothing about how work actually moves.
• The task list describes individual activities. It tells you nothing about how those activities connect, or which connections matter most.

The Workchart shows the third thing — flows. Each activity is a node; each handoff is a line; each color tells you why an activity is hard for AI to do today. The pattern that emerges is the system of work itself, and the bottlenecks are visible.

The strategic claim: instead of asking "what tasks can I automate?", you ask "what's binding throughput?" — a different and more useful question. Bottlenecks move when AI lands; tasks don't.

What you're looking at

Each circle is a unit of work — a discrete activity people in an occupation actually do.

Each line is a handoff — the output of one activity feeds the input of another. Lines marked with dashed orange highlight handoffs where a human currently has to stitch the gap. These are the openings where AI can land.

Color classifies each activity using Sangeet Paul Choudary's constraint framework from Reshuffle (2025). Three classes — what makes an activity hard for AI to absorb today:

• Scarcity (needs specialized expertise) — the activity requires deep knowledge that's hard to acquire and rare to find. AI tends to erode scarcity constraints over time.
• Risk (mistakes carry consequences) — the activity is regulated, liable, or safety-critical. AI doesn't replace human accountability here; it amplifies the value of getting it right.
• Coordination (many people need to align) — the activity involves coordinating across humans, teams, or systems. AI relocates coordination constraints — they don't disappear; they shift to new locations as automation reshapes the workflow.

Size reflects economic stakes — bigger circles mark activities where more value is at risk or in play.

The data underneath

The skeleton of the Workchart comes from publicly available data on what work looks like in the modern economy. Three sources do most of the work:

• O*NET — the US Department of Labor's catalog of occupations. ~1,000 distinct occupations defined by ~20,000 specific tasks (the activities you see as circles). O*NET also catalogs the skills, knowledge, and work activities each occupation requires. Maintained continuously by the Department of Labor since 1998. onetonline.org
• Lightcast — a labor-market data provider with billions of job postings (the demand signal — what employers actually pay for) and hundreds of millions of professional profiles (the supply signal — what skills exist where). Refreshed continuously. r.Potential ingests Lightcast via the Global Labor Graph.
• BLS (Bureau of Labor Statistics) — wage data, employment counts, and industry breakdowns by NAICS code. Tells us how big each pocket of the economy actually is.

This combination is unusual: O*NET tells you what people do; Lightcast tells you where the work is going; BLS tells you how much of it there is. None of them alone is enough; the three together give the Workchart its base layer.

Importantly, this base layer is generic — it describes the shape of work across the economy. The constraints (the colors) become specific only when a company's own context is overlaid. A company-specific Workchart blends the public skeleton with that company's actual operating reality.

How to read each view

Three kinds of views, each answering a different question:

• By Work archetype — generic priors for a category like legal drafting or clinical documentation. Useful for orienting; constraints are typical, not measured. No specific company.
• By company — a specific firm's operating model, inferred at first from public sources (annual reports, employee counts, business lines) and refined when their own data lands. Constraints reflect their reality.
• By UoP (Unit of Potential) — one alignment decision a company has made, with measured outcomes. This is where the Workchart earns its keep: you see the constraint signature before, during, and after AI lands, and the measured value that resulted.

Within an archetype or company view, the density control lets you zoom in on one role, look at a department, or pull back to the whole organization — same data, different viewpoint. A UoP already has its density baked in (it picked one when it was authored); UoP views replace density with a time control showing before / during / after states.

Status

Early access. The Workchart catalog is small at launch — one available kind-of-work view and several coming soon. Company-specific and decision-scoped views land as they're built and validated. New entries appear as the catalog grows.